The “stacking percentages” forecast, as I call it, is basically determining a bunch of small edges, small percentages, and stacking them on top of each other to create a traditional black box type model. It’s fairly simplistic but accumulates a lot of relevant factors to achieve its results. Remember, I make many lines and some are made from pieces of the stacking forecast to gauge where the market is with certain factors in play.
For these forecasts a rating is generated for each team and then a predicted score calculated. Regressions are used and the goal is to get a low line error. The goal is try to get a more realistic score than the bookmaker can put out. One test of the stacking forecast is to compare its difference from the final score to the difference between opening and closing lines and the final score.
Say the stacking forecast predicts Pittsburgh to win by 7 points and the market is showing Pittsburgh -3. If Pittsburgh wins by 10 points the market was off of reality by 7 points. The forecast was off by 3 points. In this case, the forecast was closer to reality than the line. In my opinion, crossing key numbers can affect this error so the difference between 7 and 10 is relatively smaller than 3 points when compared to the line crossing through 4, 6, and 7 in its error. In that case the book ends up being off by more than 7 points, relatively speaking. Does that make sense?
The goal of the stacking forecast is to match or beat the market error at any given time. Having a lower line error is not necessary to beat the market. Remember, when I track the stacking forecast in threads, I use as little as half a point discrepancy against the lines. I track every play, not just triggered plays (like the sharp forecast in the NFL).
But we don’t bet every game and a subset of games is usually looked at, not every game. I am tough on myself when tracking at SBR, in the CFL thread the sharp forecast is also put up against a half point discrepancy. You can test the line error at any interval or margin predicted, not just a half point difference. This is where I mentioned in the past that a larger margin doesn’t always equal a better win rate. There’s a cap and it represents a danger to teams whose ratings are pushing highs for the season.
Think tank this shit.